Photovoltaic panel defect detection model enhanced by multiscale convolution and cooperative attention
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Addressing the issues of multi-type defect recognition and low detection accuracy for small target defects in photovoltaic panel defect detection, an efficient photovoltaic panel defect detection model (TGU-YOLO) based on YOLOv11n is proposed. Firstly, multiple patterns of input features are captured through a triple parallel lightweight convolutional module to enhance the detection performance of the model. Secondly, a GAM-SimAM collaborative attention mechanism is incorporated into the neck network to achieve multi-scale enhancement of photovoltaic panel defect features and improve the feature extraction ability for small target defects. Finally, the Unified-IoU loss function is used to dynamically adjust the weights of prediction boxes, improving the robustness of target defect detection. A public photovoltaic panel defect dataset is employed for systematic evaluation, and ablation experiments, comparative experiments, and robustness tests are conducted. Experimental results show that the accuracy, recall rate, and mAP@0.5 of the TGU-YOLO model are improved by 5.4%, 7.1%, and 3.4% respectively compared to the YOLOv11n model. While maintaining a low computational load, the model achieves high-precision detection of multiple types and small target defects on photovoltaic panels, and its robustness and generalization ability are significantly enhanced.
Fields of Study & Subjects
Industrial Vision Systems and Defect Detection
Advanced Neural Network Applications
Infrastructure Maintenance and Monitoring
Publication details
- Journal / source
- Engineering Research Express
- Volume
- 8
- Issue
- 7
- Pages
- 075234–075234
- Publication date
- 1 April 2026
- Language
- EN